R 助力产品可靠性分析

make it Easier and Efficient!

刘心广

xinguang_liu@163.com

2016-12-11 @R-SHA

Outline

  • Brief Introduction of Product Reliability Engineering
  • Reliability tasks
  • What R can help, Case study
    • Modeling
    • Reporting
    • Application

What they say

First glance

Introduction

Reliability is the probability that the system will perform its required function under specified conditions for a specified period of time.

This depends on a number of factors:

  • design of the system
  • the components used
  • the environment and
  • the usage profile

Reliability Engineering is a systems approach to planning for, designing in, verifying, and tracking the reliability of products throughout whole lifecycle to achieve reliability goals.

Probability Density View

System View

Weak points

Striving to be

Features of Reliability Data

Y-Dist & Y ~ Xs

Weibull Distribution

Modeling Fit by MLE

#> Fitting of the distribution ' weibull ' by maximum likelihood 
#> Parameters : 
#>       estimate Std. Error
#> shape      3.8      0.889
#> scale     48.4      4.274
#> Loglikelihood:  -39.3   AIC:  82.6   BIC:  83.2 
#> Correlation matrix:
#>       shape scale
#> shape 1.000 0.333
#> scale 0.333 1.000

Do evaluation on vendor A:

  • Failure percentage at t = 16K is 0.015
  • Lifetime at 5% failure is 22.144

In a similar way, evaluation on vendor B:

  • Failure percentage at t = 16K is 0.088
  • Lifetime at 5% failure is 11.65

So, A is better according to requirement.

Case 1: Degradation – Crack Size Growth

Raw data plot

Modeling Fit

Regression analysis

The time at which 50% cracks would reach 1.6 inches = 0.126 millions cycles

Multi-Modeing Comparison

So if you can transfer a real problem into a mathematical problem, like regression or category, then R can do it in an easy way!

Case 2: Report in batch mode

Background: during new product qualification, lots of data generated every day but the raw record are in a mess. So the Ctr + C and Ctr + V doesn’t work here.

How to do data import, transformation, and tidy which can be used for further modeling study, and finially generate reports in batch automatically.

Demo script

R looks good

But…

Case 3: Shiny Application

Background: routing sampling test under certain stress before shipment required for condition A, where some were extended to condition B.

Data collecting in every week, but what did the data indicate from reliability view?

Demo App

Summary

R 语言博大精深

Keep learning